FLAIR: An Easy-to-Use Framework for State-of-the-Art NLP

By Alan Akbik, Tanja Bergmann, Duncan A. J. Blythe, Kashif Rasul, Stefan Schweter, Roland Vollgraf. 2019

933 citations recorded by the source corpus.

We present FLAIR, an NLP framework designed to facilitate training and distribution of state-of-the-art sequence labeling, text classification and language models. The core idea of the framework is to present a simple, unified interface for conceptually very different types of word and document embeddings. This effectively hides all embedding-specific engineering complexity and allows researchers to “mix and match” various embeddings with little effort. The framework also implements standard model training and hyperparameter selection routines, as well as a data fetching module that can download publicly available NLP datasets and convert them into data structures for quick set up of experiments. Finally, FLAIR also ships with a “model zoo” of pre-trained models to allow researchers to use state-of-the-art NLP models in their applications. This paper gives an overview of the framework and its functionality. The framework is available on GitHub at https://github.com/zalandoresearch/flair .

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

HuggingFace's Transformers: State-of-the-art Natural Language Processing (2019, neutral): “The ﬁrst direction includes Natural Language Processing libraries such as AllenNLP 13 (Gardner et al., 2018), SpaCy 14 (Honnibal and Montani, 2017), flair 15 (Akbik et al., 2018, 2019) or PyText 16 .”
ProoFVer: Natural Logic Theorem Proving for Fact Verification (2021, methodology): “We first obtain a phrase level chunking of the claims using the flair based phrase chunker (Akbik et al., 2019).”
Autoregressive Entity Retrieval (2020, methodology): “In Table 3 we compare GENRE against all methods reported in the public leaderboard: DPR (Karpukhin et al., 2020), DPR+BERT (Devlin et al., 2019), DPR+BART, tf-idf (Leskovec et al., 2014), RAG (Lewis et al., 2020), and BLINK+ﬂair (Wu et al., 2019; Akbik et al., 2019).”
KILT: a Benchmark for Knowledge Intensive Language Tasks (2020, methodology): “…sparse vector space retrieval model; (ii) DPR (Karpukhin et al., 2020)—a modern passage-level retrieval solution using dense rep-18 resentations; (iii) A combination of BLINK (Wu et al., 2019) and ﬂair (Akbik et al., 2019)—retrieval solution that ranks pages according to entities in the input.”
WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022, methodology): “To investigate whether this is also characteristic of W A NLI, we use flair (Akbik et al., 2019) to perform named entity recognition on MultiNLI and W A NLI.”
HuggingFace's Transformers: State-of-the-art Natural Language Processing (2019, neutral): “, 2018), flair (Akbik et al., 2019), and Stanza (Qi et al.”
Man is to Person as Woman is to Location: Measuring Gender Bias in Named Entity Recognition (2019, methodology): “We used Flair (Akbik, Blythe, and Vollgraf 2018; Akbik et al. 2019), CoreNLP version 3.”
Stanza: A Python Natural Language Processing Toolkit for Many Human Languages (2020, background): “While existing NLP toolkits such as CoreNLP (Manning et al., 2014), F LAIR (Akbik et al., 2019), spaCy 1 , and UDPipe (Straka, 2018) have had wide usage, they also suffer from several limitations.”
Transformers: State-of-the-Art Natural Language Processing (2020, neutral): “, 2018), flair (Akbik et al., 2019), and Stanza (Qi et al.”
Multi-step Jailbreaking Privacy Attacks on ChatGPT (2023, methodology): “We use FLAIR (Akbik et al., 2019) to extract named entities (NEs) from predicted contents and target email contents.”
TX-Ray: Quantifying and Explaining Model-Knowledge Transfer in (Un-)Supervised NLP (2019, methodology): “oup tokens activations by their word function to better read, analyze and compare feature preference distributions – see Fig. 3, 5, 7 or 9. POS tags are produced by the state-of-the-art Flair tagger (Akbik et al., 2019) using the Penn Treebank II4 tag set. 3Though possible, we do not pretrain Transf”
Low Resource Multi-Task Sequence Tagging - Revisiting Dynamic Conditional Random Fields (2020, methodology): “Our implementation extends the popular sequence labeling framework FLAIR1 (Akbik et al., 2019).”
German’s Next Language Model (2020, methodology): “For GermEval14 we report the result that can be achieved using the FLAIR framework (Akbik et al., 2019).”
Pooled Contextualized Embeddings for Named Entity Recognition (2019, methodology): “Our reproduced numbers are slightly lower than we reported in Akbik et al. (2018) where we used the official CONLL-03 evaluation script over BILOES tagged entities.”
ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic (2020, background): “context-sensitive (hence more accurate) representations of words, and a growing list of embeddings (Akbik et al., 2019) followed.”